File size: 16,812 Bytes
95f8c4b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 | """Gemini-compatible application service handlers."""
from __future__ import annotations
import json
from fastapi import HTTPException
from fastapi.responses import StreamingResponse
from aistudio_api.api.response_models import (
GeminiCandidateResponse,
GeminiContentResponse,
GeminiGenerateContentResponse,
)
from aistudio_api.api.responses import to_gemini_parts, to_gemini_usage_metadata
from aistudio_api.api.schemas import GeminiGenerateContentRequest
from aistudio_api.api.state import runtime_state
from aistudio_api.application.api_service_common import (
MAX_RETRIES,
ensure_active_account,
logger,
record_rotator_event,
require_busy_lock,
try_switch_account,
)
from aistudio_api.application.chat_service import (
cleanup_files,
normalize_gemini_request,
)
from aistudio_api.domain.errors import (
AistudioError,
AuthError,
RequestError,
UsageLimitExceeded,
)
from aistudio_api.infrastructure.gateway.client import AIStudioClient
async def handle_gemini_generate_content(
model_path: str,
req: GeminiGenerateContentRequest,
client: AIStudioClient,
*,
stream: bool,
):
busy_lock = require_busy_lock()
if stream:
return _build_gemini_streaming_response(
client=client, req=req, model_path=model_path
)
last_error = None
for attempt in range(MAX_RETRIES):
async with busy_lock:
await ensure_active_account(attempt)
normalized = None
try:
normalized = normalize_gemini_request(req, model_path)
logger.info(
"Gemini: model=%s, contents=%s, stream=False, attempt=%d",
normalized.model,
len(req.contents),
attempt + 1,
)
output = await client.generate_content(
model=normalized.model,
capture_prompt=normalized.capture_prompt,
capture_images=normalized.capture_images,
contents=normalized.contents,
system_instruction_content=normalized.system_instruction,
tools=normalized.tools,
safety_settings=normalized.safety_settings,
temperature=normalized.temperature,
top_p=normalized.top_p,
top_k=normalized.top_k,
max_tokens=normalized.max_tokens,
generation_config_overrides=normalized.generation_config_overrides,
sanitize_plain_text=False,
)
record_rotator_event("success")
runtime_state.record(normalized.model, "success", output.usage)
return GeminiGenerateContentResponse(
candidates=[
GeminiCandidateResponse(
content=GeminiContentResponse(
parts=to_gemini_parts(
output.text,
function_calls=output.function_calls,
function_responses=output.function_responses,
thinking=output.thinking,
images=output.images,
reasoning_images=output.reasoning_images,
),
),
finishReason="STOP"
if not output.function_calls
else "FUNCTION_CALL",
)
],
usageMetadata=to_gemini_usage_metadata(output.usage),
)
except ValueError as exc:
raise HTTPException(
400, detail={"message": str(exc), "type": "bad_request"}
) from exc
except UsageLimitExceeded as exc:
runtime_state.record(model_path, "rate_limited")
last_error = exc
record_rotator_event("rate_limited")
if await try_switch_account():
logger.info(
"Gemini 429 限流，已切换账号，重试 %d/%d",
attempt + 1,
MAX_RETRIES,
)
continue
logger.warning("Gemini 429 限流，无法切换账号")
raise HTTPException(
429, detail={"message": str(exc), "type": "rate_limit_exceeded"}
) from exc
except AistudioError as exc:
runtime_state.record(model_path, "errors")
record_rotator_event("error")
raise HTTPException(
500, detail={"message": str(exc), "type": "server_error"}
) from exc
except Exception as exc:
runtime_state.record(model_path, "errors")
record_rotator_event("error")
logger.error("Gemini error: %s", exc, exc_info=True)
raise HTTPException(
500, detail={"message": str(exc), "type": "server_error"}
) from exc
finally:
if normalized is not None and not stream:
cleanup_files(normalized.cleanup_paths)
raise HTTPException(
429, detail={"message": str(last_error), "type": "rate_limit_exceeded"}
) from last_error
def _build_gemini_streaming_response(
*,
client: AIStudioClient,
req: GeminiGenerateContentRequest,
model_path: str,
) -> StreamingResponse:
async def stream_response():
busy_lock = runtime_state.busy_lock
if busy_lock is None:
yield (
"data: "
+ json.dumps(
{"error": {"message": "Server not ready"}}, ensure_ascii=False
)
+ "\n\n"
)
return
async with busy_lock:
normalized = None
try:
await ensure_active_account(0)
normalized = normalize_gemini_request(req, model_path)
logger.info(
"Gemini stream: model=%s, contents=%s",
normalized.model,
len(req.contents),
)
final_usage = None
has_yielded_data = False
for stream_attempt in range(MAX_RETRIES):
try:
has_yielded_data = False
async for event_type, text in client.stream_generate_content(
model=normalized.model,
capture_prompt=normalized.capture_prompt,
capture_images=normalized.capture_images,
contents=normalized.contents,
system_instruction_content=normalized.system_instruction,
tools=normalized.tools,
safety_settings=normalized.safety_settings,
temperature=normalized.temperature,
top_p=normalized.top_p,
top_k=normalized.top_k,
max_tokens=normalized.max_tokens,
generation_config_overrides=normalized.generation_config_overrides,
sanitize_plain_text=False,
force_refresh_capture=stream_attempt > 0,
):
has_yielded_data = True
if event_type == "body" and text:
yield (
"data: "
+ json.dumps(
{
"candidates": [
{
"content": {
"role": "model",
"parts": [{"text": text}],
},
"finishReason": None,
}
]
},
ensure_ascii=False,
)
+ "\n\n"
)
elif event_type == "images" and text:
img_list = text if isinstance(text, list) else []
yield (
"data: "
+ json.dumps(
{
"candidates": [
{
"content": {
"role": "model",
"parts": [
part.model_dump(
mode="json",
exclude_none=True,
)
for part in to_gemini_parts(
"", images=img_list
)
],
},
"finishReason": None,
}
]
},
ensure_ascii=False,
)
+ "\n\n"
)
elif event_type == "reasoning_images" and text:
r_img_list = text if isinstance(text, list) else []
yield (
"data: "
+ json.dumps(
{
"candidates": [
{
"content": {
"role": "model",
"parts": [
part.model_dump(
mode="json",
exclude_none=True,
)
for part in to_gemini_parts(
"",
reasoning_images=r_img_list,
)
],
},
"finishReason": None,
}
]
},
ensure_ascii=False,
)
+ "\n\n"
)
elif event_type == "thinking" and text:
yield (
"data: "
+ json.dumps(
{
"candidates": [
{
"content": {
"role": "model",
"parts": [
{
"text": text,
"thought": True,
}
],
},
"finishReason": None,
}
]
},
ensure_ascii=False,
)
+ "\n\n"
)
elif event_type == "usage":
final_usage = text if isinstance(text, dict) else None
break
except UsageLimitExceeded:
if normalized is not None:
runtime_state.record(normalized.model, "rate_limited")
record_rotator_event("rate_limited")
if not has_yielded_data and await try_switch_account():
logger.warning(
"Gemini stream 429 限流，已切换账号，重试 %d/%d",
stream_attempt + 1,
MAX_RETRIES,
)
continue
raise
except RequestError as exc:
if exc.status == 204 and stream_attempt == 0:
logger.warning(
"Gemini stream 收到 204，清理 snapshot 缓存后重试一次"
)
client.clear_snapshot_cache()
continue
raise
except AuthError as exc:
if stream_attempt == 0:
logger.warning(
"Gemini stream 鉴权异常，清理 snapshot 缓存后重试一次: %s",
exc,
)
client.clear_snapshot_cache()
continue
raise
record_rotator_event("success")
if normalized is not None:
runtime_state.record(normalized.model, "success", final_usage)
if final_usage:
yield (
"data: "
+ json.dumps(
{
"candidates": [],
"usageMetadata": to_gemini_usage_metadata(
final_usage
).model_dump(mode="json"),
},
ensure_ascii=False,
)
+ "\n\n"
)
yield "data: [DONE]\n\n"
except Exception as exc:
logger.error("Gemini stream error: %s", exc, exc_info=True)
if not isinstance(exc, UsageLimitExceeded):
record_rotator_event("error")
if normalized is not None:
runtime_state.record(normalized.model, "errors")
yield (
"data: "
+ json.dumps({"error": {"message": str(exc)}}, ensure_ascii=False)
+ "\n\n"
)
finally:
if normalized is not None:
cleanup_files(normalized.cleanup_paths)
return StreamingResponse(
stream_response(),
media_type="text/event-stream",
headers={"Cache-Control": "no-cache", "X-Accel-Buffering": "no"},
)
|